TSDatasets.__init__ regressors logic

[julia-shenshina]

🚀 Feature Request

Now TSDataset handles regressors/exogs according to theirs column name prefixes. In this issue we want to fix it: now user should add info about regressor nature of data in df_exog.

Motivation

New way of work with regressors.

Proposal

• Add TSDataset.__init__ argument known_future:

def __init__(self, df: pd.DataFrame, freq: str, df_exog: Optional[pd.DataFrame] = None, known_future: Optional[List[str]] = None):
        """Init TSDataset.
        Parameters
        ----------
        df:
            dataframe with timeseries
        freq:
            frequency of timestamp in df
        df_exog:
            dataframe with exogenous data;
            if the series is known in the future features' names should start with prefix 'regressor_`.
        known_future: 
            series from columns in df_exog[known_future] are regressors; if None given,  all the given in df_exog series are not regressors
        """

• Replace _update_regressors in __init__ with new logic:
  • now only columns in known_future should be added to self._regressors
• Change logic in _check_regressors:
  • now we don't have any prefixes and work only with columns in self._regressor

Test cases

Update existing cases with new logic

Alternatives

No response

Additional context

No response

Checklist

• I discussed this issue with ETNA Team